Multivariate analysis of solar city economics: Impact of …...risk, and reduce project costs.7,8 It...

Advanced Review

Multivariate analysis of solar cityeconomics: impact of energyprices, policy, finance, and coston urban photovoltaic powerplant implementationJohn Byrne ,1,2 Job Taminiau,1,2* Kyung Nam Kim,3* Joohee Lee 1 andJeongseok Seo 1

Previous research suggests that the potential for city-scale photovoltaic (PV)applications is substantial across the globe. Successful implementation of ‘solarcity’ options will depend on the strategic application of finance mechanisms,solar energy soft cost policies, and other policy tools, as well as the grid price ofelectricity. Capital markets recently have embraced the roll-out of new financialinstruments, including ‘green bonds,’ which could be incorporated into solar cityproject design to attract large investments at a low cost. A multivariate analysismethod is employed to consider solar city possibilities for six cities: Amsterdam,London, Munich, New York, Seoul, and Tokyo. A Monte Carlo simulation is con-ducted to capture the probabilistic nature of uncertainties in the parameters andtheir relative importance to the financial viability of a solar city project. The anal-ysis finds that solar city implementation strategies can be practical under a broadrange of circumstances. © 2017 John Wiley & Sons, Ltd

How to cite this article:WIREs Energy Environ 2017, 6:e241. doi: 10.1002/wene.241

INTRODUCTION

The December 2015 Paris Climate Agreement sig-nals a new direction for global climate change.For the first time, commitments by nearly all coun-tries have been agreed upon. While the scale ofrequired change sizably surpasses these commit-ments, the Paris Agreement has the distinct advan-tage of directing attention to implementation ratherthan policy design. In this vein, a critical challenge is

the immaturity of the renewable energy sector as atarget for city-scale development. Investing in renew-able energy still focuses mainly on individual, mod-estly sized projects. For instance, solar electric powertypically attracts investments in a few hundred kWpto tens of MWp. A growing number of policy ana-lysts and technology researchers argue that a newfocus is needed on infrastructure-scale planning toadvance low carbon energy transitions.1,2

At the same time, experience shows that infra-structure investment must respond to context- andlocation-specific factors, leading many to emphasize apost-Paris ‘polycentricity paradigm’ in climate changegovernance and economics where policy, planning,and strategy is operationalized by nation–state,regional, city, and city network initiatives.3 Cities, inparticular, have positioned themselves as championsof sustainable energy change, combining their effortsin transnational networks and collaboration.4

*Correspondence to: [email protected]; [email protected] for Energy and Environmental Policy, University of Dela-ware, Newark, DE, USA2Foundation for Renewable Energy & Environment, New York,NY, USA3Green School, Korea University, 145, Anam-ro, Seoul, South Korea

Conflict of interest: The authors have declared no conflicts of inter-est for this article.

Volume 6, Ju ly /August 2017 © 2017 John Wiley & Sons, Ltd 1 of 16

http://orcid.org/0000-0001-5626-0950http://orcid.org/0000-0002-5883-8362http://orcid.org/0000-0003-0584-0394

To that end, this article explores the possibilityfor cities to engage in network-wide climate changestrategies to attract capital market attention throughcapitalization of existing and currently unutilizedrooftop real estate that is abundant in the urbanenvironment. Termed ‘solar cities,’ such strategiesuse city governance vehicles to catalyze publicsector-led, infrastructure-scale design and investmentof city-wide photovoltaic (PV) technology deploy-ment. Successful access to capital markets couldincrease the scale and speed of solar deployment.5

For instance, city-scale implementation efforts couldbenefit from the rising prominence of ‘green bonds,’‘climate bonds,’ and other financing innovationsthat have been able to expand large-scale capitalinvestments for green purposes.2 Global issuance asof August 2016 (i.e., 9 months of investments)stands at $46.03 billion, or $5 billion more thaninvestments in all of 2015.a The emerging financialinstrument could be a promising vehicle for solarcity strategies: incorporating public debt in thefinancing structure offers attractive benefits such asimprovements in financing terms, risk mitigation,and access to a broader capital pool.6 Overall, suchstrategies could achieve a lower cost of capital, lessrisk, and reduce project costs.7,8 It is estimated thata reduction in the levelized cost of energy (LCOE)of 8–16% occurs when a portion of a project orportfolio’s typical capital stack is replaced with pub-lic capital vehicles.5

This study advances earlier work from Ref 2by introducing a multivariate analysis method forassessing solar city economics in the same six citiesstudied in the previous publication: Amsterdam,London, Munich, New York City, Seoul, andTokyo. The analysis is organized in five steps. Afterreviewing newly published work in the field, whichpoints to significant untapped potential (BriefReview of Recent ‘Solar City’ Assessment Literaturesection), the paper describes a multivariate methodto evaluate solar city economics (Analytic Approachsection). The next section covers a project financeanalysis of data to illustrate the relative impacts ofmarket, finance, and policy portfolios in solar cityeconomics (Project Finance Analysis section). Usingregression analysis, the variables driving solar cityviability are assessed (Regression Results section),and the model is evaluated using statistical tests forrobustness and validity (Model Robustness Testssection). Conclusion section concludes that city-scaleapplications are practical—the necessary policy toolsexist and have been shown to work, and the eco-nomics and financeability of such projects areaffordable.

BRIEF REVIEW OF RECENT ‘SOLARCITY’ ASSESSMENT LITERATURE

The latest modeling assessments and other researchcontinue to strengthen the case for urban ‘solar city’applications, where urban energy economies areretooled toward a strong reliance on PV energy genera-tion using the large rooftop area available to cities.2,9

Research regarding urban applications of PV hasaddressed, with growing sophistication, methodologi-cal issues in determining the overall technical potentialof PV in cities.10,11 Such investigations have been per-formed for a wide variety of urban conditions rangingfrom cities in Nigeria12 to India,13 Abu Dhabi,14 theNetherlands,15 the United States,16,17 and Brazil.18

Studies typically find considerable potential forurban deployment of PV. For example, Gagnonet al.16 calculate that the suitable rooftop space inthe state of California can generate 74% of the elec-tricity sold by utilities in 2013, while several NewEngland states are found to be able to generate over45% of their electricity needs by utilizing existingrooftop area. At the national level, Gagnon et al.16

estimate that rooftop solar power could generate38.6% of national electricity demand, and similarly,at the city level, cities like Los Angeles (60% of elec-tricity needs), San Francisco (50%), Miami (46%),and Atlanta (41%) show substantial potential.b Gag-non et al.16 performed their calculation for 47 citiesacross the United States and found that, collectively,these cities have the technical potential to host animpressive 84.4 GWp of solar capacity—the SolarEnergy Industry Association (SEIA) puts the currenttotal U.S.-installed PV capacity at about 25 GWp.

19

Los Angeles (9 GWp), New York City (8.6 GWp),and Chicago (6.9 GWp) combined could match cur-rent U.S.-installed capacity.

Other dimensions of solar energy potential mod-eling have been explored by looking at, e.g., a varietyof rooftop technologies,20 different configurationsof urban morphology,21,22 varying system designconsiderations,23 optimization opportunities,24 or pos-sible interaction patterns with mobility options.25

The practical implementation of solar cities asan adaptive strategy available to local policymakers, however, remains limited. While there isvalue in further discussion and research regardingmethodology refinement, research targeting thepracticality of the solar city concept as a city-widestrategy is timely and necessary to advance the field.An earlier attempt to do so introduced the policy,market, and finance implications associated withsolar city strategies.2

Advanced Review wires.wiley.com/energy

2 of 16 © 2017 John Wiley & Sons, Ltd Volume 6, July/August 2017

ANALYTIC APPROACH

The examination of solar city’s practicality relies onthree analytical tools: (1) project finance analysis,(2) regression analysis, and (3) robustness tests. Eachis described in this section and then applied in thefollowing sections.

Benefit–Cost AnalysisA first assessment of project viability is to determinethe benefit cost ratio in the first year of the project(Eq. (1)) and the benefit cost ratio in the year thedebt matures (Eq. (6)). Calculations and analysiswere performed using the System Advisor Model(SAM) software developed at the National Renewa-ble Energy Laboratory (NREL) in the United States.All dollar amounts are in nominal dollars.

BCn =VF+PBOM +DS

ð1Þ

where n = year of the project. The other variablesgiven in Eq. 1 are further defined in Eq. 2-5.

Value factor VFð Þ=OF× 1−DRð Þn ×ER× 1+EGRð Þn $½ � ð2Þ

The output factor (OF) in kWh/kWp refers to thecombined effect of the typical meteorological yearoutput data that is specific to the meteorology of thegeographical location, PV system characteristics(e.g., module efficiency, tilt, etc.), and morphologicalconditions of the city in question. See Table 1 for theOF of each location.

The degradation rate (DR), in %, accounts forsystem losses over time and is set at 0.5% across alllocations following Ref 26.

The electricity rate (ER), in cents/kWh, is theadministratively set commercial ER that is applicablein the case study city as an approximate measure ofthe value of electricity generated by the PV powerplant.c See Table 1 for the ER of each location.

The electricity growth rate (EGR), in %, reflectsthe expected growth pattern of the ER. The EGR isset at 2% in each location.

Policy benefits PBð Þ =VF× Policy support $½ � ð3Þ

Policy support refers to the existing policy conditionsto support PV generation in each city, including thenational policy system, which are described in detailin Ref 2. Policy support value is assumed to be inplace for 10 years, after which it is eliminated. Policysupport values are presented in Table 1 for eachlocation.

Operations andmaintenance OMð Þ= constant× 1 + ið Þn $½ � ð4Þ

Following Ref 2, the OM constant was set at $25.0/kWp,

d and inflation (i) is set at 2% for each location.

Debt service DSð Þ= PMT IR, MAT, HCð Þ+ PMT IR, MAT, SCð Þ $½ � ð5Þ

Debt service (DS) refers to the constant periodic pay-ment required to pay off the capital investment, both

TABLE 1 | Overview of the Input Data

CityOutput Factor(kWh/kWp)

Electricity RetailRate ($/kWh)

PolicyBenefits ($/kWh)

HardCosts ($/kW)

SoftCosts ($/kW)

Capital Costs (InterestRate) (%)1

AMS 967 0.148 0.114 1498 642 1.8

LON 979 0.167 0.16 1530 1020 3.6

MUN 1079 0.208 —2 1435 615 1.5

NYC 1364 0.224 0.0993 2046 1674 3.1

SEOUL 1110 0.116 0.127 1470 980 3.5

TOKYO 1218 0.194 0.077 2154 1436 0.6

Explanation about data and data sources provided in Byrne et al.21 Capital costs given in Table 1 are for a 10-year bond offering. Interest rates differ per bond maturity; relationship of change in the form of yield curves isprovided by Byrne et al.2

2 Recent changes to the German renewable energy support structure are an example of the uncertainty developers and investors can face. Policy changes toreduce feed-in tariff payments are one factor contributing to a recent decline in German PV installations from 2013’s 3.3 GWp to 1.4 GWp of new capacityin 2015. This installation level is down from the 7.5 GWp installed annually in 2010, 2011, and 2012.31,79 Of course, other factors played a role in thisdecline, but the policy decision (as part of a ‘third phase’ in the German ‘Energiewende’) to recognize rapidly falling technology costs and a maturing marketwas important. The third phase of implementation reflects the view that ‘grid parity’ is in sight, and efficient deployment of solar PV can be achieved at lowerlevels of policy support.80 Munich’s policy support is set at zero by the authors in light of the recent decision by the government to significantly reduce itsincentive programs. In this way, the results reported here reflect a robust range of scenarios where significant policy support is present and where policy sup-port has been removed.

3 For New York City, the capital-based incentive is calculated as a production-based compensation for a 10-year period for comparison.

WIREs Energy and Environment Multivariate analysis of solar city economics


‘soft’ and ‘hard’ costs (see below), with a constantinterest rate (IR) over a specified period (MAT). DSis calculated using EXCEL’s PMT function that com-bines IR, principal (HC and SC), and matu-rity (MAT).e

Byrne et al.2 previously considered project via-bility in a relatively straightforward manner: solarcities were assessed to be viable if annual benefits(derived from both policy support and energy sales)were higher than annual costs (consisting of invest-ment, operation and maintenance, and interest) forall years of the project’s lifetime (assumed to be25 years). An important variable over the lifetime ofthe project is the duration of policy support: policybenefits (PBs) typically expire before the end of thetechnical lifetime of the project. This variable mustrely on an assumption regarding their expiration. Inthe 2016 article by Byrne et al., a 10-year retirementwas assumed and is repeated here. However, thistreatment can cause annual costs to be higher thanannual benefits for a short period of time right afterPB expiration, threatening project viability. Despitesubstantial positive cash flows for the first 10 yearsof city projects, long financing maturities wererequired to overcome this obstacle. However, suchshortfalls can be brief and small compared to the sur-plus generated in the first 10 years of the projects.For this reason, project viability is redefined in thisanalysis to be based on cumulative cash flow: cityprojects are viable if cumulative cash flow is positivethroughout the lifetime of the project. Using the sameterms as above, but including a cost and price escala-tor of 2% and PV system performance degradationof 0.5%, the cumulative benefit-to-cost ratio deter-mines the project viability for a 25-year PV installa-tion (Eq. (6)).

Project viability = IF BCyr1 ≥ 0ANDBCy1 +BCyr2 ≥ 0�

ANDBCyr1 + BCyr2 +… +BCyr25 ≥ 0Þ ð6Þ

Key model input variables that are subject to varia-bility tests (described below) are provided in Table 1for each city. Variability in investment conditions cansubstantially alter the risk profile of renewableenergy:

• Policy changes. Policy support uncertainty canreduce investor confidence and limitinvestment.27–30 Recent examples of such policyuncertainty are the retroactive modification ofSpain’s feed-in tariff and Germany’s substantiallowering of its price premium.29,31

• Financing period. In response to a ‘mock’ solarsecuritization filing, U.S. rating agencies indi-cated that the typical 20-year contract lifetimefor PV projects is unlikely to remain as the mar-ket matures and proposed 7- to 10-year maturi-ties as a more reasonable timeframe for analysispurposes.32

• Cost of capital. The absence of performancedata regarding infrastructure-scale PV securiti-zations fuels a cyclical phenomenon where ‘riskperception is fed by lack of historical knowl-edge, which is in turn fed by risk perception.’33

As such, ‘it may be realistic to assume that thefirst securitizations will not obtain an optimalspread between the cost of capital of securitiza-tion debt and, say, that of a commercial loan.’6

Credit enhancement techniques, such as over-collateralization, first-loss reserves, or tranchingcan, on the other hand contribute to loweringthe cost of capital.6,33

• PV system output. The output function of asolar PV system is dependent on a wide rangeof underlying variables. For instance, systemperformance is determined by meteorologicalconditions, urban morphology, module effi-ciency, angular deployment of PV panels, sys-tem performance degradation, PV technologychoice, and shading (e.g., from other rooftopobstacles, from other buildings, or panel-to-panel shading). Urban morphology conditionsare particularly different across locations andrange from macro-scale to micro-scale elements:city districts (e.g., comparing a business districtagainst a suburban neighborhood) provide dif-ferent shading conditions, different heating/cooling requirements, or are subject to differentfire codes, and similarly, individual buildingscan be designed specifically with PV in mind orbe ornate and largely unsuitable for PVdeployment.34–36 By relying on the modificationof six variables, such as the amount of openspace, the closeness of buildings, and the stand-ard deviation of building heights, certain build-ing design changes could improve London’srooftop PV performance by about 9% whileenhancing façade profiles by 45%.35 A study ofNew York’s Battery Park commercial district(a vertical, high-density area) shows annualrooftop solar irradiation variability from 1200to 1350 kWh/m2—or about 6.4–16.8% lowercompared to the unobstructed annual irradia-tion level.37



• Technology and market dynamics. Solar electricpower generation technologies have experienceda dramatic drop-off in overall costs.38 However,while median prices of solar power have stead-ily declined, significant variability in the cost ofboth rooftop and commercial-scale solar instal-lations remains. For instance, in the UnitedStates, the price point difference between the20th percentile and 80th percentile is $1.7/Wpfor residential and $1.6/Wp–$1.3/Wp for non-residential systems.38 Moreover, dominant mar-ket relations, geographic market conditions,and contractual arrangements influence the pat-tern of technology price development (e.g., byaffecting feedstock trade flows or costs).39 Priceand cost volatility characterizes the learningcurve of PV.40 Also, as prices come down forsome components, other components or aspectsbecome more prominent in future prices.41,42 Aparticular distinction can be made between the‘hard’ and ‘soft’ cost patterns of PV, and a dis-cussion on this balance in each location is pro-vided below.

Hard Costs Versus Soft CostsOverall, ‘hard’ costs of solar electric power, particu-larly PV module prices, have fallen dramatically.38

For instance, module prices fell by $2.7/Wp (2014dollars) over the 2008–2012 period.38 Nonmodulecost improvements, however, have also contributedto a continuing decline in installed costs.43 For exam-ple, since 2009, nonmodule costs have decreased by10% year-over-year and contributed to a $0.4/Wdecline in cost from 2013 to 2014.38 Key contribu-tors to these price reduction patterns are lower costsfor inverters and racking equipment and falling aver-age generation costs due to increasing system sizeand module efficiency.38,44 However, increasing scru-tiny directed toward ‘soft’ costs, such as permitting,regulatory context, marketing, customer acquisition,installer margins, installation labor, and systemdesign, has likely also contributed to installed costreductions for PV. 21f

Recent research into PV-installed costs hasfocused on so-called ‘soft costs,’ e.g., Refs 46–52.Considering hardware costs are ’fairly similar’ acrosscountries, soft cost profiles can be a key differentiatorin installed costs.53 Soft cost ‘best practices’ can sig-nificantly reduce overall costs, e.g., Refs 49. Impor-tantly, soft costs in the German system are about50% lower than the United States,49 in large partbecause the former has standardized both the designapproval and permitting processes. For instance, the

difference between the United States and Germany isstriking:

the feed-in tariff registration form, which enablesgrid-connected solar residences to receive federalincentives, is the only German paperwork requiredfor PV systems. Typically, this form takes as little as5 min to fill out and is conveniently submittedonline. In contrast, most U.S. [jurisdictions] require acombination of engineering drawings, building per-mit, electrical permit, design reviews, and multipleinspections before approving a PV installation(Ref 41).

Differences in soft costs occur within, as well asacross, countries. A recent analysis of U.S. soft costsfound an 8–12% reduction potential for a highest-scoring municipality compared to the lowest-scoringsoft cost community.51 Similarly, Dong and Wiserfind that the most favorable permitting practices incities in California resulted in reductions of $0.27/Wp–$0.77/Wp lower costs (4–12% of median Cali-fornia price points) compared to cities with the leastfavorable permitting practices.54 Researchers havealso found that regulatory and policy barriers thataffect soft costs in some cases would lead installers toavoid certain jurisdictions.55,56

Establishing specific soft costs at the city levelrequires an investigation into factors such as permit-ting, installation labor, margins, etc that is beyond thescope of the present research. Instead, we focus on theratio between soft and hard costs for each city asdetermined by extant literature. National-level dataare used where available and extrapolated to neigh-boring countries where local data are unavailable.

Soft costs have been extensively studied in theUnited States, e.g., Refs 46–51,57. The results pointto a substantial potential in installation cost reduc-tion as soft costs make up a significant share of totalinstalled costs. For instance, in the first half of 2012,soft costs made up 64% of the total U.S. residentialinstalled cost, 57% of U.S. small commercial, and52% of U.S. large commercial projects.55 Similarresults were produced in a more recent investigation:55% of residential costs were attributable to softcosts, commercial projects devote 42% of installationcosts to soft costs, and utility-scale projects mustcover 34–36% in soft costs.47 In the analysis pre-sented here, we assume that New York City’s instal-lation cost is made up of 45% soft costs and 55%hard costs. This assumption is similar to the ratio forcommercial projects in the United States, a reasona-ble facsimile for our solar city buildout onNew York’s rooftops.



The extant literature commonly points toGermany as an example of best practices in terms ofsoft costs.41,58 Research has been particularly direc-ted toward the soft cost profiles of the residentialmarket. For instance, Seel et al.49 compare soft costsbetween Germany and the United States and find thatsoft cost differences for residential systems amountedto $2.72/W in 2011–2012. This means a doubling ofsoft cost payments in the United States—Germanyspends 21% of a residential system installation($0.62/W) on soft costs compared to 54% in theUnited States ($3.34/W).49 A separate analysis by theGerman solar energy association finds that just over31% of a commercial application of rooftop solarenergy can be attributed to soft costs, and a recentstudy by the Fraunhofer Institute uses 30% for thiscost element.59g For these reasons, we assume a 70%module and hard cost and 30% soft cost profile forthe City of Munich.

A recent publication found a similar differencein cost between Japan and the United States.48 Hardcosts, particularly module costs, are $0.67/W higherin Japan compared to the United States, leading to adifferent soft cost/hard cost balance: soft costsaccount for about 44% of residential and 39% ofsmall commercial system costs in the first half of2013.48 This finding is further supported by datafrom the International Energy Agency (IEA) Photo-voltaic Power Systems Programme (PVPS) forJapan.60 The inputs for Tokyo used here are, there-fore, that soft cost prices make up 40% of systemprice, while hard costs make up 60%. This putsTokyo in between Germany and the United States interms of the proportion of installed costs dedicatedto the soft, downstream elements of the value chain.

Detailed soft cost versus hard cost breakdownsfor other countries remain to be investigated. Weassume here that Amsterdam has a similar profile toMunich,h London has a similar profile as New YorkCity,i and Seoul has a similar profile as Tokyo.j Theinputs for soft cost versus hard cost are provided inTable 2.

Monte Carlo SimulationBecause full-scale solar city implementations have yetto occur, a Monte Carlo simulation approach wasused to test the impact of variability and to illustratethe consequences of uncertainty. Probabilistic analy-sis techniques represent a suitable tool for thistask.66–68 Monte Carlo simulation is a well-recognized method, and the analysis below parallelsits application performed in other studies, e.g., Refs66,69,70. Input variable sampling is carried out by

using a random number generator within a prese-lected input variability range. Repeating the proce-dure for a large number of simulations generates adensity function for the model output values. Thecalculated output (see Benefit-Cost Analysis section)is conducted for each input variable separately tosimulate the possible solar city economics faced bylocal governments in the selected locations. Theapproach is illustrated in Figure 1.

To test solar city financeability more robustly,the Monte Carlo assessment was performed for�5%, �10%, and �15% changes in the input vari-ables. The input variables selected to be subject tothis range of change were: (1) bond capital costs,(2) the electricity retail price at the moment of bondissue for which solar generated electricity can be sold,(3) the rate of change in electricity retail price overtime, (4) the ‘hard’ and ‘soft’ costs of the PV system,and (5) the level of policy support (see below for sta-tistical justification of the selected variables). Theanalysis was conducted across bond issue maturitiesfrom 6 to 16 years, with 10,000 simulations createdfor each variability range (i.e., �5%, �10%, and�15%), each maturity, and for each city to ensurerandom selection of benefit-to-cost ratios. This matu-rity range was selected to encompass the suggestedtimeframe by rating agencies of 7–10 years (see Ref32) and include the high end of financial feasibilityuncovered by Ref 2. Across three ranges ofvariability—6 cities and 11 maturities—the MonteCarlo assessment yields a total of just under 2 millionsimulations.

Regression AnalysisOur next step was to build a multiple regressionmodel to determine the ability of each input factor to

TABLE 2 | Inputs for Hard and Soft Cost Percentage for Each Cityin the Analysis

City Hard Costs (%) Soft Costs (%)

Low soft costs

Amsterdam 70 30

Munich 70 30

Medium soft costs

Seoul 60 40

Tokyo 60 40

High soft costs

New York City 55 45

London 55 45

Sources: Based on national data reported by Refs 25,38,41,47,57 and extra-polations where national data are unavailable.



predict solar city benefit–cost ratios. The analysisranks cost and benefit inputs according to the magni-tude and sign of standardized regression coefficients(SRCs).71k This statistical approach allows for thequantification of predictors and a ranking of theirrelative importance in determining cumulativebenefit-to-cost ratios. A regression analysis was con-ducted for each city individually, for each maturity,and for all cities combined for each maturity. Weused EViews Version 8.1 software to run the calcula-tions (http://www.eviews.com/).

The independent variables used for the analysiswere: (1) the IR charged on the bond offering, (2) thehard costs of the installation (HC), (3) the soft costsof the PV installation (SC), (4) the OF of the system(OF), (5) the PBs offered in each location (PB),(6) the growth rate of the retail electricity pricethroughout the bond offering (i.e., the price escala-tor; EGR), and (7) the commercial retail rate in eachlocation (ER). Monte Carlo variations in each ofthese model inputs were used to determine thebenefit-to-cost ratio in the year of maturity (i.e., atthe conclusion of the financing round) as the depend-ent variable (see Eq. (1)). The relationship isdescribed in Eq. (7), with an error term u.

BC =B1IR +B2HC+B3SC+B4OF+B5PB+B6EGR+B7ER+u ð7Þ

Standardization of the regression is described inEq. (8). Standard errors for the βs in Eq. (8) arebased on White’s heteroscedasticity-consistent

standard error formula72 using the function providedby the Eviews Version 8.1 software.

BCstd = β1IR + β2HC+ β3SC + β4OF+ β5PB + β6EGR+ β7ER+u ð8Þ

Model Robustness CheckGraphical model validation checks were carried outto evaluate the performance of the model. In particu-lar, the distribution of the residual errors offersinsight into the functioning of the regression model.In addition, using three different variability ranges(�5, �10, and �15%) adds to the understanding ofthe robustness of the model.

PROJECT FINANCE ANALYSIS

Two cities—New York and Munich—have finan-cially viable solar city investment opportunitiesregardless of variability in input values (Figure 2). Asdepicted in Figure 2, illustrating the comprehensiveresults of the Monte Carlo assessment with combinedvariability in input data of � 5%, � 10%, and�15%, New York, for any maturity greater than12 years, can confidently expect a benefit-to-costratio greater than 1.0. In Munich’s case, this point isreached for maturities equal to or greater than13 years. For New York and Munich, 90% of thesimulations for the combined variability scenarios

Model input (x1,2,...,n)ranges

5% – x1,2,...,n + 5%Model inputs (x1,2,...,n)per city per maturity

Model output y

Probabilitydistribuitionfunction of y

Benefit-to-cost ratiooutputscalculated bydeterministicmodel for ~2 millionsimulations

Random numbergenerator

(10,000× per range)

x1,2,...,n(30,000observationsper variable)

Deterministic model*(financeability of solar city project)

10% – x1,2,...,n + 10%

15% – x1,2,...,n + 15%

FIGURE 1 | Monte Carlo assessment of project finance. *NREL’s System Advisor Model (SAM) is used to calculate the benefit–cost ratios(https://sam.nrel.gov/).



http://www.eviews.com/https://sam.nrel.gov/

show ratios greater than 1.0 from, respectively,12 and 13 years onward.

In a second tier of investment opportunities,Amsterdam, London, and Tokyo need longer to real-ize financeability; all appear financeable with maturi-ties of 16 years. The cumulative benefits of solar cityinvestments in the three cities exceed, or are veryclose in exceeding, all costs in 90% or more of thesimulations when the financing can be stretched over16 years. Amsterdam, London, and Tokyo reach amean benefit-to-cost ratio larger than 1.0 under 11-,12-, and 13-year financing maturities, respectively.

The third tier, consisting of the City of Seoul,cannot confidently offer net positive investment oppor-tunities (defined as the point at which 90% of thesimulations have cumulative net positive cash flows)when cost input variability can be as high as +15%.Seoul’s solar city investment achieves positive cash

flow only for a small share of the 30,000 simulationsunder 16-year financing conditions. As discussedbelow, a key barrier is its grid price, which the nationalgovernment sets administratively. The current price is3.2–10.8 cents lower than its five counterparts.

REGRESSION RESULTS

The analysis was conducted for both each city indi-vidually for each maturity and for all cities combinedfor each maturity.

Regressions by CityWe were able to identify predictor structures for eachof the six cities. These are reported in Table 3. ForNew York, the high electricity retail rate delivers asubstantial contribution to the financial viability of its

FIGURE 2 | Results of the Monte Carlo analysis using a 90% interval around the mean for each city for a combined variability of the inputdata. Notes: Combined variability is achieved by varying input data by �5% for 10,000 simulations per maturity year, �10% for 10,000simulations per maturity year, and �15% variability for another 10,000 simulations per maturity year. A total of 30,000 simulations, therefore, areassessed to determine the benefit–cost ratios. The mean and 90% range correspond with the left y-axis, and the columns correspond with theright y-axis, depicting the percentage of simulations that are defined as viable projects (i.e., positive cumulative benefit cost ratios in all years ofthe project, using Eq. (6)). The distinctive ‘bend’ in the results is a direct effect of the assumed 10-year lifespan of the policy benefits in eachlocation: as soon as the policy benefits expire, the benefit-to-cost ratio relies solely on the retail electricity rate and electricity growth rate todetermine the benefits component of the analysis. A gradual phase-out of the policy benefits or other mitigating strategies could shorten therequired financing timeframe.



solar city project. In contrast, the low electricity retailrate in Seoul (the lowest in the sample of cities) is thekey impediment to financial viability of the project.

Other findings include:

• Borrowing IRs have minor influence on finan-cial viability compared to the other factors. Thisfinding holds across all locations. However, interms of policy strategies and options toincrease the viability of a solar city project,developers might have some control over thecost of finance, e.g., by focusing on a sociallyresponsible or ‘impact’ investor base.

• Price developments throughout the projectappear to have a modest effect on project viabil-ity. Unsurprisingly, changing the price point atwhich the generated electricity is sold has a sub-stantial effect on PV competitiveness.

• The relative influence of soft costs in Amster-dam and Munich, two cities where substantialprogress on soft cost improvements has alreadytaken place, is also modest. The relative influ-ence of the variable is more pronounced insome of the other locations.

Overall, predictors had the expected signs, and acommon predictor structure can be uncovered. Usingstandardized heteroscedasticity robust coefficients todefine variable contribution, the system OF variablehas the greatest predictive impact on project financialviability. Retail ERs are the second most influentialfactor in contributing to positive benefit–cost ratios.Solar panel costs (HC) have the largest negative influ-ence on project cash flows. Although its predictivepower varies due to differences among the cities and

the national commitments that underpin local efforts,the importance of PBs is evident. The common pre-dictive structure and variation in influence of thevariables is captured in Figure 3.

Regression Analyzing the Combined Poolof CitiesConducting the regression analysis across the pool ofsix cities shows that the electricity retail rate is thelargest factor affecting positive cash flow, with solarradiation a very close second factor. Because of largedifferences in city/national SC policies, it ranks as amore prominent factor compared to solar panel costsin shaping overall project cost (Table 4).

The regression analysis across the pool of sixcities further allows for the calculation of the relativevalue of PBs across locations (Table 5). The viabilityof London and Seoul’s solar projects are compara-tively more dependent on policies in place in theirmarkets, while the projects of Amsterdam, New YorkCity, and Tokyo are less dependent. The third columnin the table offers a brief explanation of the relativeordering of the cities.

RESULT OF THE MODEL ROBUSTNESSCHECK

The very large sample sizes provided by the MonteCarlo simulation (180,000 simulations in the analysisof combined samples) make the typical normalitytests unhelpful here as these tests are intended forassessments involving smaller sample sizes (typicallyless than 50 cases). Instead, model validation wasconsidered by using graphical tools. A histogram eye-ball test suggests a normal distribution of analysis

TABLE 3 | Overview of Regression Results by City

Variable

Standardized Regression Coefficient

Amsterdam London Munich New York Seoul Tokyo

Interest rate −0.0591 −0.1184 −0.0484 −0.0860 −0.1133 −0.0211

Hard costs −0.4428 −0.3933 −0.3955 −0.3040 −0.3890 −0.3842

Soft costs −0.1888 −0.2628 −0.1694 −0.2510 −0.2611 −0.2561

Output factor 0.7023 0.7088 0.6343 0.6055 0.7066 0.6839

Policy benefit 0.2753 0.3132 — 0.3032 0.3346 0.1685

Electricity growth rate 0.0402 0.0362 0.0578 0.0555 0.0348 0.0466

Electricity rate 0.4287 0.3994 0.6344 0.6077 0.3758 0.5151

R2 0.9954 0.9954 0.9947 0.9937 0.9955 0.9951

Results are provided for a 10-year maturity schedule. Interest rate and electricity growth rate variables show the most change with changes in maturity sched-ule (longer maturity schedules make these variables more important). The policy benefit variable shows a higher level of change after 10-year maturity (benefitsare assumed to expire after 10 years). Yet, the relative position of variables remains regardless of maturity schedule. Germany has substantially lowered its pol-icy incentive programs, and as a result, Munich’s regression was run without this variable.



FIGURE 3 | Multivariate regression results for each city for the seven variables. Notes: For each city, the ranking of the coefficients shows theinfluence of that variable on the benefit–cost ratio. Generation potential (kWh/m2) and the electricity retail rate are the key drivers of benefit–costratio outcomes in most locations. A notable exception is the City of Seoul where the electricity retail rate is relatively less important due to its lowstarting point.



residuals (Figure 4). Other graphical assessments,such as normal probability plots and quantile–quantile plots, offer a similar outcome. These resultslead to the finding of normality in the model’s distri-bution of residuals.

CONCLUSION

Solar cities, as a technical option, have now beenfirmly established in the literature and represent apromising total deployment potential for urban energyeconomy restructuring.2,10,16,73 The question remains,however, if they are practical. The Paris Agreementcommitments will require significant financing and

access to the capital markets, including a rapid ramp-up of ‘green bonds,’ ‘climate bonds,’ or other innova-tive capital financing structures.2,74,75 Moving beyondmethodological approaches to determine solar citypotential, the article explores possible pathways for theactual implementation of solar cities using capital mar-ket finance strategies that are consistent with the emer-ging green bond and climate bond markets. Theanalysis reported here illustrates that solar cities couldbe practical infrastructure-scale development strategiesfor local policy makers. Particularly for New YorkCity and Munich, even accounting for possible fluctua-tion in insolation, technology cost, policy benefits, andfuture grid prices, a solar city option appears feasible.Indeed, sensitivity analyses performed for this articleindicate that all six cities could establish a relativelylow-cost policy regime to finance solar cities; the Cityof Seoul, however, would require timeframes longerthan 16 years for debt repayment. Ongoing PV marketmaturation and integration, combined with ongoingadvancements in PV efficiency, reliability, cost profiles,manufacturing, and business model reform, will likelyimprove the overall attractiveness of the solar cityoption.

Benefit and cost considerations associated withPV deployment are many.76 A range of risks andpotential opportunities, such as security of supplyand the technical, environmental, or social implica-tions of widespread deployment of solar PV in theurban environment, are not included in the presentanalysis and remain open for future study. For exam-ple, no effort was made to credit solar city projectswith the value of improved air quality in or beyondtheir borders. Another consideration excluded in thecurrent analysis is the option for cities to work

TABLE 5 | Overview of the PB in Each Location

CityPB StandardizedRegression Coefficient Comparative PBs

Munich — No PB needed, PVmarket ready

Tokyo 0.7328 Low PB, high relianceon market factors

New YorkCity

1.1069 Capital PB, highelectricity price

Amsterdam 1.1434 Moderate PB,moderate electricityprice

Seoul 1.2574 Considerable PB, lowelectricity price

London 1.4258 High PB, moderateelectricity price

PB, policy benefit.Results are provided for a 10-year maturity schedule. The PB variable showsa higher level of change after 10-year maturity (benefits are assumed toexpire after 10 years). Yet, the relative position of variables remains regard-less of maturity schedule.

TABLE 4 | Overview of the regression results of the combinedregression analysis.

Variable Standardized Regression Coefficient

IR −0.4083

HC −0.7950

SC −0.9381

EGR 0.0334

OF 1.0677

ER 1.3070

EGR, electricity growth rate; ER, electricity rate; HC, hard costs; IR, interestrate; OF, output factor; SC, soft costs.Numbers provided here are for a 10-year maturity schedule of the debt.Interest rate and electricity growth rate variables show the most change withchanges in maturity schedule (longer maturity schedules make these vari-ables more important). Yet, the relative position of variables remainsregardless of maturity schedule. FIGURE 4 | Histogram overview of model residuals for all cities

combined for a 10-year maturity. Black line is a normal distributionoverlay.



together in creating solar cities. Cities have been veryactive in policy networks, for instance, to address cli-mate change and other environmental challenges.Information exchanges have also been a focus ofurban policy networks aimed at spreading ‘best prac-tice’ strategies to improve sustainability.l

An example of how networks could be fruitfullyexplored concerns SC profiles. Amsterdam andMunichhave brought down their PV SC profiles significantly.Sharing their experiences and approach with othercities could be an important means for bringing downSCs elsewhere. In addition, cities could band togetherto bring down other hard costs through pooled pro-curement models and financing structures. Other toolsexcluded in the present analysis are the option toinclude additional technology options in the financing(especially energy efficiency options),77,78 to hybridizethe bond or other debt offerings, or to include optionsto manage project pipelines by offloading previouslyinstalled solar city components through the actions ofsecondary market participants, such as yieldcos.m

In sum, research needs regarding solar citiesremain large, but the promise of the option wouldappear to merit increased attention.

NOTESa Several organizations monitor the growth of the greenbond market. Numbers here are taken from the ClimateBonds Initiative (CBI; https://www.climatebonds.net/).b It is important to note here that Gagnon et al.16 calculategeneration as a percentage of total consumption over thecourse of an entire year. The percentage of consumptionthat could be generated for daylight energy needs or forseasonal needs could be substantially higher. In previouslypublished research, Byrne et al. found in one city—Seoul—that the technical generation potential of a rooftop powerplant exceeds the entire electricity demand of the city dur-ing the hours of 11 a.m.–2 p.m. in May.9c The commercial electricity retail rate is used as anapproximate measure of the value of electricity generatedin order to reflect the idea that the city’s PV generation sys-tem can be seen to operate as one aggregate unit. All elec-tricity produced is used within the city’s limits, and ineffect, all electricity generated by the distributed PV powerplant is available for self-consumption.d The operating and maintenance costs (O&M) are notincluded in the variability tests. As was performed in Ref 2,the O&M costs are taken to be the same in each case studylocation (25 $/kWp/year). Sources to support this assump-tion are provided in Ref 2.e This is an Excel financial function that calculates the con-stant payment required to pay off a constant IR debt for a

specified period. In this case, the maturity is the specifiedperiod, and the IR is determined from the associated yieldcurve.f Sizable reductions in technology prices can result innations adopting protectionist policies, such as increasedimport duties and local content requirements, to shieldtheir local markets from global market competition. Suchapproaches could deter investment and hamper down-stream implementation and job creation.45

g The German solar energy association published its pricemonitor findings in early 2013. Their database does notappear to include more recent publications on the pricecomponents of solar energy in Germany. However, a sim-plified price breakdown illustrated by the Fraunhofer Insti-tute31 that shows little movement in the ratio of moduleand nonmodule costs for recent years suggests the hard ver-sus soft cost balance identified by the German solar energyassociation is a useful approximation.h Considering both the Netherlands’ slight outperformancecompared to Germany in the early 2000s in terms of softcosts,61 but also subsequent Dutch instability in the PVmarket due to frequent policy changes,62 this assumptionhas some empirical backing. This assumption finds furthersupport in data from the IEA PVPS: recent communicationsfrom the Netherlands have consistently indicated a lowshare of nonmodule costs in system prices.60i Recent work by the London Assembly Environment Com-mittee highlights how nonsolar-specific factors, like highparking costs and the city’s congestion charge, could hindersolar energy deployment in the city.63 This notion of a ’has-sle factor’ is also reported in grey literature publications,such as a 2015 article by the Guardian.64 In addition, theLondon Assembly report points to New York City as acomparable case that should be consulted in London’sefforts to expand the use of PV.j Data on Seoul’s soft versus hard costs of PV installation islimited. Considering the similarities between Korea andJapan’s energy development pathway, the hard-to-soft costbalance empirically observed for Tokyo can be considereda suitable assumption for Seoul.65k SRCs remove differences in units of measurement fordependent and independent variables, thereby enabling acomparison of the relative effects of predictors on the pre-dicted variable without regard to scale differences in varia-ble measurement.l A list of networks performing these functions wouldalways include ICLEI, C40, the European Climate Alliance,and the U.S. Conference of Mayors Climate ProtectionAgreement.m Yieldcos allow for offloading previously installed solarcity components through the actions of secondary marketparticipants. Yieldcos are dividend growth-oriented entitiesthat, in this case, would bundle renewable long-term con-tracted operating assets in order to generate predictablecash flows attractive to yield-focused investors.



https://www.climatebonds.net/

ACKNOWLEDGMENTS

This work was supported by the National Research Foundation of Korea Grant funded by the Korean Govern-ment (Ministry of Science, ICT & Future Planning) (2015, University-Institute Cooperation Program). Thismaterial is based upon work primarily supported by the National Science Foundation (NSF) and the Depart-ment of Energy (DOE) under NSF CA No. EEC 1041895. Any opinions, findings and conclusions or recom-mendations expressed in this material are those of the author(s) and do not necessarily reflect those of NSFor DOE.

[Correction added on 16 May 2017, after first online publication: additional funding information has beeninserted in the Acknowledgements section.]

REFERENCES1. Hall S, Foxon T, Bolton R. Investing in low-carbon

transitions: energy finance as an adaptive market. ClimPolicy 2015. doi:10.1080/14693062.2015.1094731.

2. Byrne J, Taminiau J, Kim K, Seo J, Lee J. A solar citystrategy applied to six municipalities: integrating mar-ket, finance, and policy factors for infrastructure-scaledevelopment in Amsterdam, London, Munich,New York, Seoul, and Tokyo. Wiley InterdiscipRev Energy Environ 2016, 5:68–88. doi:10.1002/wene.182.

3. Jordan AJ, Huitema D, Hildén M, van Asselt H,Rayner TL, Schoenefeld JJ, Tosun J, Forster J,Boasson EL. Emergence of polycentric climate govern-ance and its future prospects. Nat Clim Change 2015,5:977–982. doi:10.1038/nclimate2725.

4. Kern K, Bulkeley H. Cities, Europeanization andmulti-level governance: governing climate changethrough transnational municipal networks. J CommonMarket Stud 2009, 47:309–332. doi:10.1111/j.1468-5965.2009.00806.x.

5. Mendelsohn M, Feldman D. Financing U.S. RenewableEnergy Projects through Public Capital Vehicles: Quali-tative and Quantitative Benefits. Golden, CO: NationalRenewable Energy Laboratory (NREL); 2013.

6. Lowder T, Mendelsohn M. The Potential of Securitiza-tion in Solar PV Finance. Golden, CO: NationalRenewable Energy Laboratory; 2013.

7. Lowder T, Schwabe P, Zhou E, Arent D. Historicaland Current U.S. Strategies for Boosting DistributedGeneration. Golden, CO: National Renewable EnergyLaboratory; 2015.

8. Alafita T, Pearce J. Securitization of residential solarphotovoltaic assets: costs, risks and uncertainty.Energy Policy 2014, 67:488–498. doi:10.1016/j.enpol.2013.12.045.

9. Byrne J, Taminiau J, Kurdgelashvili L, Kim K. Areview of the solar city concept and methods to assessrooftop solar electric potential, with an illustrativeapplication to the city of Seoul. Renew Sustain EnergyRev 2015, 41:830–844. doi:10.1016/j.rser.2014.08.023.

10. Freitas S, Catita C, Redweik P, Brito M. Modelingsolar potential in the urban environment: state-of-the-art review. Renew Sustain Energy Rev 2015,41:915–931. doi:10.1016/j.rser.2014.08.060.

11. Martin A, Dominguez J, Amador J. Applying LIDARdatasets and GIS based model to evaluate solar poten-tial over roofs. AIMS Energy 2015, 3:326–343.doi:10.3934/energy.2015.3.326.

12. Okoye C, Taylan O, Baker D. Solar energy potentialsin strategically located cities in Nigeria: review,resource assessment and PV system design. Renew Sus-tain Energy Rev 2016, 55:550–566. doi:10.1016/j.rser.2015.10.154.

13. Singh R, Banerjee R. Estimation of rooftop solar pho-tovoltaic potential of a city. Solar Energy 2015,115:589–602. doi:10.1016/j.solener.2015.03.016.

14. Joshi B, Hayk B, Al-Hinai A, Woon W. Rooftop detec-tion for planning of solar PV deployment: a case studyin Abu Dhabi. In: Woon W, Aung Z, Madnick S, eds.Data Analytics for Renewable Energy Integration,vol. 8817. Cham, Switzerland: Springer InternationalPublishing; 2014, 137–149. doi:10.1007/978-3-319-13290-7_11.

15. Kausika B, Dolla O, Folkerts W, Siebenga B,Hermans P, van Sark W. Bottom-up analysis of thesolar photovoltaic potential for a city in theNetherlands—a working model for calculating thepotential using high resolution LiDAR data. In: 2015International Conference on Smart Cities and GreenICT Systems (SMARTGREENS). Lisbon: IEEE; 2015,1–7. Available at: http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=7297968&tag=1. (Accessed Decem-ber 1, 2016)

16. Gagnon P, Margolis R, Melius J, Phillips C, Elmore R.Rooftop Solar Photovoltaic Technical Potential in theUnited States: A Detailed Assessment. Golden, CO:National Renewable Energy Laboratory; 2016.

17. Kurdgelashvili L, Li J, Shih C-H, Attia B. Estimatingtechnical potential for rooftop photovoltaics in Cali-fornia, Arizona, and New Jersey. Renew Energy 2016,95:286–302. doi:10.1016/j.renene.2016.03.105.



http://dx.doi.org/10.1080/14693062.2015.1094731http://dx.doi.org/10.1002/wene.182http://dx.doi.org/10.1002/wene.182http://dx.doi.org/10.1038/nclimate2725http://dx.doi.org/10.1111/j.1468-5965.2009.00806.xhttp://dx.doi.org/10.1111/j.1468-5965.2009.00806.xhttp://dx.doi.org/10.1016/j.enpol.2013.12.045http://dx.doi.org/10.1016/j.enpol.2013.12.045http://dx.doi.org/10.1016/j.rser.2014.08.023http://dx.doi.org/10.1016/j.rser.2014.08.023http://dx.doi.org/10.1016/j.rser.2014.08.060http://dx.doi.org/10.3934/energy.2015.3.326http://dx.doi.org/10.1016/j.rser.2015.10.154http://dx.doi.org/10.1016/j.rser.2015.10.154http://dx.doi.org/10.1016/j.solener.2015.03.016http://dx.doi.org/10.1007/978-3-319-13290-7_11http://dx.doi.org/10.1007/978-3-319-13290-7_11http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=7297968&tag=1http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=7297968&tag=1http://dx.doi.org/10.1016/j.renene.2016.03.105

18. Miranda R, Szklo A, Schaeffer R. Technical-economicpotential of PV systems on Brazilian rooftops. RenewEnergy 2015, 75:694–713. doi:10.1016/j.renene.2014.10.037.

19. Kann S, Shiao M, Honeyman C, Jones J, Perea A,Smith C. Solar Market Insight 2015 Q3. Washington,DC: SEIA and GTM Research; 2015. Available at:http://www.seia.org/research-resources/solar-market-insight-2015-q3. (Accessed December 1, 2016).

20. Rosas-Flores J, Rosas-Flores D, Zayas J. Potentialenergy saving in urban and rural households of Mex-ico by use of solar water heaters, using geographicalinformation systems. Renew Sustain Energy Rev 2016,53:243–252. doi:10.1016/j.rser.2015.07.202.

21. Takebayashi H, Ishii E, Moriyama M, Sakaki A,Nakajuma S, Ueda H. Study to examine the potentialfor solar energy utilization based on the relationshipbetween urban morphology and solar radiation gainon building rooftops and wall surfaces. SolarEnergy 2015, 119:362–369. doi:10.1016/j.solener.2015.05.039.

22. Sarralde J, Quinn D, Wiesmann D, Steemers K. Solarenergy and urban morphology: scenarios for increasingthe renewable energy potential of neighborhoods inLondon. Renew Energy 2015, 73:10–17. doi:10.1016/j.renene.2014.06.028.

23. Lund P, Mikkola J, Ypyä J. Smart energy systemdesign for large clean power schemes in urban areas.J Clean Prod 2015, 103:437–445. doi:10.1016/j.jclepro.2014.06.005.

24. Li X, Wen J, Malkawi A. An operation optimizationand decision framework for a building cluster with dis-tributed energy systems. Appl Energy 2016,178:98–109. doi:10.1016/j.apenergy.2016.06.030.

25. Mendoza J-MF, Sanyé-Mengual E, Angrill S, García-Lozano R, Feijoo G, Josa A, Gabarrell X,Rieradevall J. Development of urban solar infrastruc-ture to support low-carbon mobility. Energy Policy2015, 85:102–114. doi:10.1016/j.enpol.2015.05.022.

26. Jordan DC, Kurtz SR. Photovoltaic degradationrates—an analytical review. Prog Photovolt 2013,21:12–29. doi:10.1002/pip.1182.

27. White W, Lunnan A, Nybakk E, Kulisic B. The role ofgovernments in renewable energy: the importance ofpolicy consistency. Biomass Bioenergy 2013,57:97–105. doi:10.1016/j.biombioe.2012.12.035.

28. Polzin F, Migendt M, Täube FA, von Flotow P. Publicpolicy influence on renewable energy investments—apanel data study across OECD countries. EnergyPolicy 2015, 80:98–111. doi:10.1016/j.enpol.2015.01.026.

29. De la Hoz J, Martin H, Miret J, Castilla M,Guzman R. Evaluating the 2014 retroactive regulatoryframework applied to the grid connected PV systemsin Spain. Appl Energy 2016, 170:329–344.doi:10.1016/j.apenergy.2016.02.092.

30. Mignon I, Bergek A. Investments in renewable electric-ity: the importance of policy revisited. Renew Energy2016, 88:307–316. doi:10.1016/j.renene.2015.11.045.

31. Fraunhofer Institute for Solar Energy Systems. Recentfacts about photovoltaics in Germany: market seriesreport on the German photovoltaic industry; 2016.Available at: https://www.ise.fraunhofer.de/ (AccessedApril 22, 2016).

32. Mendelsohn M, Lowder T, Rottman M, Borod R,Gabig N, Henne S, Caplin C, Notte Q. The SolarAccess to Public Capital (SAPC) Mock SecuritizationProject. Golden, CO: National Renewable Energy Lab-oratory; 2015.

33. Mendelsohn M, Udanick M, Joshi J. Credit Enhance-ments and Capital Markets to Fund Solar Deployment:Leveraging Public Funds to Open Private SectorInvestment. Golden, CO: National Renewable EnergyLaboratory; 2015.

34. Takebayashi H, Ishii E, Moriyama M, Sakaki A,Nakajuma S, Ueda H. Study to examine the potentialfor solar energy utilization based on the relationshipbetween urban morphology and solar radiation gainon building rooftops and wall surfaces. Solar Energy2015, 119:362–369.

35. Sarralde J, Quinn D, Wiesmann D, Steemers K. Solarenergy and urban morphology: scenarios for increasingthe renewable energy potential of neighborhoods inLondon. Renew Energy 2015, 73:10–17.

36. Robinson D. Urban morphology and indicators ofradiation availability. Solar Energy 2006,80:1643–1648.

37. Lee K, Lee J, Lee J. Feasibility study on the relationbetween housing density and solar accessibility andpotential uses. Renew Energy 2016, 85:749–758.doi:10.1016/j.renene.2015.06.070.

38. Barbose G, Darghouth N, Millstein D, Spears M,Wiser R, Buckley M, Widiss R, Grue N. Tracking theSun VIII - The Installed Price of Residential and Non-Residential Photovoltaic Systems in the United States.Berkeley, CA: Lawrence Berkeley National Laboratory,U.S. Department of Energy; 2015.

39. Yu Y, Li H, Bao H. Price dynamics and market rela-tions in solar photovoltaic silicon feedstock trades.Renew Energy 2016, 86:526–542. doi:10.1016/j.renene.2015.08.069.

40. Candelise C, Winskel M, Gross RJK. The dynamics ofsolar PV costs and prices as a challenge for technologyforecasting. Renew Sustain Energy Rev 2013,26:96–107. doi:10.1016/j.rser.2013.05.012.

41. Ardani K, Seif D, Margolis R, Morris J, Davidson C,Truitt S, Torbert R. Non-Hardware ("Soft") Cost-Reduction Roadmap for Residential and Small Com-mercial Solar Photovoltaics, 2013–2020. Golden, CO:National Renewable Energy Laboratory; 2013.

42. Burrows K, Fthenakis V. Glass needs for a growingphotovoltaics industry. Solar Energy Mater Solar



http://dx.doi.org/10.1016/j.renene.2014.10.037http://dx.doi.org/10.1016/j.renene.2014.10.037http://www.seia.org/research-resources/solar-market-insight-2015-q3http://www.seia.org/research-resources/solar-market-insight-2015-q3http://dx.doi.org/10.1016/j.rser.2015.07.202http://dx.doi.org/10.1016/j.solener.2015.05.039http://dx.doi.org/10.1016/j.solener.2015.05.039http://dx.doi.org/10.1016/j.renene.2014.06.028http://dx.doi.org/10.1016/j.renene.2014.06.028http://dx.doi.org/10.1016/j.jclepro.2014.06.005http://dx.doi.org/10.1016/j.jclepro.2014.06.005http://dx.doi.org/10.1016/j.apenergy.2016.06.030http://dx.doi.org/10.1016/j.enpol.2015.05.022http://dx.doi.org/10.1002/pip.1182http://dx.doi.org/10.1016/j.biombioe.2012.12.035http://dx.doi.org/10.1016/j.enpol.2015.01.026http://dx.doi.org/10.1016/j.enpol.2015.01.026http://dx.doi.org/10.1016/j.apenergy.2016.02.092http://dx.doi.org/10.1016/j.renene.2015.11.045https://www.ise.fraunhofer.de/http://dx.doi.org/10.1016/j.renene.2015.06.070http://dx.doi.org/10.1016/j.renene.2015.08.069http://dx.doi.org/10.1016/j.renene.2015.08.069http://dx.doi.org/10.1016/j.rser.2013.05.012

Cells 2015, 132:455–459. doi:10.1016/j.solmat.2014.09.028.

43. Nemet GF, O’Shaughnessy EO, Wiser R,Darghouth N, Barbose G, Gillingham K, Rai V. Char-acteristics of Low-Priced Solar Photovoltaic Systems inthe United States. Berkeley, CA: Lawrence BerkeleyNational Laboratory; 2016. Available at: https://emp.lbl.gov/. (Accessed December 1, 2016).

44. Wang X, Byrne J, Kurdgelashvili L, Barnett A. Highefficiency photovoltaics: on the way to becoming amajor electricity source. Wiley Interdiscip Rev EnergyEnviron 2012, 1:132–151. doi:10.1002/wene.44.

45. OECD. Overcoming Barriers to International Invest-ment in Clean Energy. Green Finance and Investment,OECD Publishing: Paris; 2015. doi:10.1787/9789264227064-en.

46. Friedman B, Ardani B, Feldman D, Seif D,Davidson C, Morris J, Truitt S, Torbert R,Margolis R. What’s driving the price of PV? RenewEnergy Focus 2014, 15:26–29. doi:10.1016/S1755-0084(14)70067-5.

47. Chung D, Davidson C, Fu R, Ardani K, Margolis R.U.S. Photovoltaic Prices and Cost Breakdowns: Q12015 Benchmarks for Residential, Commercial, andUtility-Scale Systems. Golden, CO: National Renewa-ble Energy Laboratory; 2015.

48. Friedman B, Margolis R, Seel J. Comparing Photovol-taic (PV) Costs and Deployment Drivers in the Japa-nese and US Residential and Commercial Markets.Golden, CO: National Renewable Energy Laboratory/Lawrence Berkeley National Laboratory; 2014.

49. Seel J, Barbose G, Wiser R. An analysis of residentialPV system price differences between the United Statesand Germany. Energy Policy 2014, 69:216–226.doi:10.1016/j.enpol.2014.02.022.

50. Ardani K, Barbose G, Margolis R, Wiser R,Feldman D, Ong S. Benchmarking Non-Hardware Bal-ance of System (Soft) Costs for U.S. Photovoltaic Sys-tems Using a Data-Driven Analysis from PV InstallerSurvey Results. Golden, CO: National RenewableEnergy Laboratory; 2012.

51. Burkhardt J, Wiser R, Dargouth N, Dong C,Huneycutt J. Exploring the impact of permitting andlocal regulatory processes on residential solar prices inthe United States. Energy Policy 2015, 78:102–112.doi:10.1016/j.enpol.2014.12.020.

52. Woodhouse M, Jones-Albertus R, Feldman D, Fu R,Horowitz K, Chung D, Jordan D, Kurtz S. The Role ofAdvancements in Solar Photovoltaic Efficiency, Relia-bility, and Costs. Golden, CO: National RenewableEnergy Laboratory (NREL); 2016.

53. Feldman D, Barbose G, Margolis R, Bolinger M,Chung D, Fu R, Seel J, Davidson C,Wiser R. Photovoltaic System Pricing Trends—Histori-cal, Recent, and Near-Term Projections. Golden, CO:U.S. Department of Energy, SunShot; 2015.

54. Dong C, Wiser R. The impact of city-level permittingprocesses on residential photovoltaic installation pricesand development times: an empirical analysis of solarsystems in California cities. Energy Policy 2013,63:531–542. doi:10.1016/j.enpol.2013.08.054.

55. Friedman B, Ardani K, Feldman D, Citron R,Margolis R, Zuboy J. Benchmarking Non-HardwareBalance-of-System (Soft) Costs for U.S. PhotovoltaicSystems, Using a Bottom-Up Approach and InstallerSurvey. Golden, CO: National Renewable Energy Lab-oratory; 2013.

56. Tong J. Nationwide Analysis of Solar Permitting andthe Implications for Soft Costs. San Francisco, CA:Clean Power Finance; 2012.

57. Fox E, Edwards T. 2015 South Carolina PV Soft Costand Workforce Development. Aiken, SC: SavannahRIver National Laboratory; 2016.

58. Morris J, Calhoun K, Goodman J, Seif D. ReducingSolar PV Soft Costs—A Focus on Installation Labor.Boulder, CO: Rocky Mountain Institute; 2013.

59. Bundesverband Solarwirtschaft (BSW-Solar). Photovol-taik Preismonitor Deutschland (German PV Module-PriceMonitor 2013); 2013. Available at: https://www.solarwirtschaft.de/preisindex. (Accessed December1, 2016).

60. IEA PVPS. Trends 2015 in PhotovoltaicApplications—Survey Report of Selected IEA Coun-tries Between 1992 and 2014. Paris: IEA PVPS; 2015.

61. Schaeffer GJ, Alsema E, Seebregts A, Beurskens L, deMoor H, van Sark W, Durstewitz M, Perrin M,Boulanger P, Laukamp H, et al. Learning from theSun: Analysis of the Use of Experience Curves forEnergy Policy Purposes: The Case of PhotovoltaicPower. Final Report of the Photex Project. Petten, TheNetherlands: Energy Centre Netherlands (ECN); 2004.

62. Huijben J, Verbong G. Breakthrough without subsi-dies? PV business model experiments in the Nether-lands. Energy Policy 2013, 56:362–370. doi:10.1016/j.enpol.2012.12.073.

63. London Assembly Environment Committee. Bring MeSunshine! How London’s Homes Could GenerateMore Solar Energy. London, UK: Greater LondonAuthority; 2015. Available at: https://www.london.gov.uk/sites/default/files/07a_environment_committee_-_domestic_solar_report_-_final.pdf. (Accessed December1, 2016).

64. Vaughan A. Why London is rubbish at solar. TheGuardian; 2015. Available at: https://www.theguardian.com/environment/2015/jan/26/why-london-is-rubbish-at-solar (Accessed January 26, 2015).

65. Chen W-M, Kim H, Yamaguchi H. Renewable energyin eastern Asia: renewable energy policy review andcomparative SWOT analysis for promoting renewableenergy in Japan, South Korea, and Taiwan. EnergyPolicy 2014, 74:319–329.



http://dx.doi.org/10.1016/j.solmat.2014.09.028http://dx.doi.org/10.1016/j.solmat.2014.09.028https://emp.lbl.gov/https://emp.lbl.gov/http://dx.doi.org/10.1002/wene.44http://dx.doi.org/10.1787/9789264227064-enhttp://dx.doi.org/10.1787/9789264227064-enhttp://dx.doi.org/10.1016/S1755-0084(14)70067-5http://dx.doi.org/10.1016/S1755-0084(14)70067-5http://dx.doi.org/10.1016/j.enpol.2014.02.022http://dx.doi.org/10.1016/j.enpol.2014.12.020http://dx.doi.org/10.1016/j.enpol.2013.08.054https://www.solarwirtschaft.de/preisindexhttps://www.solarwirtschaft.de/preisindexhttp://dx.doi.org/10.1016/j.enpol.2012.12.073http://dx.doi.org/10.1016/j.enpol.2012.12.073https://www.london.gov.uk/sites/default/files/07a_environment_committee_-_domestic_solar_report_-_final.pdfhttps://www.london.gov.uk/sites/default/files/07a_environment_committee_-_domestic_solar_report_-_final.pdfhttps://www.london.gov.uk/sites/default/files/07a_environment_committee_-_domestic_solar_report_-_final.pdfhttps://www.theguardian.com/environment/2015/jan/26/why-london-is-rubbish-at-solarhttps://www.theguardian.com/environment/2015/jan/26/why-london-is-rubbish-at-solarhttps://www.theguardian.com/environment/2015/jan/26/why-london-is-rubbish-at-solar

66. Arnold U, Yildiz Ö. Economic risk analysis of decen-tralized renewable energy infrastructures—a MonteCarlo simulation approach. Renew Energy 2015,88:227–239. doi:10.1016/j.renene.2014.11.059.

67. da Silva Pereira E, Pinho J, Galhardo M, Macêdo W.Methodology of risk analysis by Monte Carlo methodapplied to power generation with renewable energy.Renew Energy 2014, 69:347–355. doi:10.1016/j.renene.2014.03.054.

68. United States Environmental Protection Agency. Guid-ing principles for Monte Carlo Analysis. Washington,DC: United States Environmental Protection Agency;1997. Available at: http://www.epa.gov/risk/guiding-principles-monte-carlo-analysis. (Accessed December1, 2016).

69. Bianchini F, Hewage K. Probabilistic social cost-benefitanalysis for green roofs: a lifecycle approach. BuildEnviron 2012, 58:152–162. doi:10.1016/j.buildenv.2012.07.005.

70. Glatzmaier G, Gomez J. Determining the cost benefitof high-temperature coatings for concentrating solarpower thermal storage using probabilistic cost analysis.J Solar Energy Eng 2015, 137:041006-1–041006-7.doi:10.1115/1.4029862.

71. Helton J, Davis F. Sampling-Based Methods for Uncer-tainty and Sensitivity Analysis. Albuquerque, NM:Sandia National Laboratories; 2000.

72. White H. A heteroskedasticity-consistent covariancematrix estimator and a direct test for heteroskedasticity.Econometrica 1980, 48:817–838. doi:10.2307/1912934.

73. Castillo CP, Batista e Silva F, Lavalle C. An assessmentof the regional potential for solar power generation inEU-28. Energy Policy 2016, 88:86–99. doi:10.1016/j.enpol.2015.10.004.

74. Covington H. Investment consequences of the Paris cli-mate agreement. J Sustain Finan Invest 2016.doi:10.1080/20430795.2016.1196556.

75. Ng TH, Tao JY. Bond financing for renewable energyin Asia. Energy Policy 2016, 95:509–517.doi:10.1016/j.enpol.2016.03.015.

76. Hansen L, Lacy V, Glick D. A Review of Solar PVBenefit & Cost Studies. 2nd ed. Boulder, CO: RockyMountain Institute; 2013.

77. Rosaria C, Francesco M, Lorenzo A, Mario P. SolarStreet lighting: a key technology en route to sustaina-bility. Wiley Interdiscip Rev Energy Environ 2016.doi:10.1002/wene.218.

78. Laitner JA. Linking energy efficiency to economic pro-ductivity: recommendations for improving the robust-ness of the U.S. economy. Wiley Interdiscip RevEnergy Environ 2015, 4:235–252. doi:10.1002/wene.135.

79. Fraunhofer Institute for Solar Energy Systems.Photovoltaics report: market series report on the(global) photovoltaic industry; 2016. Availableat: https://www.ise.fraunhofer.de/ (Accessed June6, 2016).

80. Fulton M, Capalino R, Auer J. The German Feed-inTariff: Recent Policy Changes. Frankfurt am Main:Deutsche Bank; 2012. Available at: www.dbresearch.com.



http://dx.doi.org/10.1016/j.renene.2014.11.059http://dx.doi.org/10.1016/j.renene.2014.03.054http://dx.doi.org/10.1016/j.renene.2014.03.054http://www.epa.gov/risk/guiding-principles-monte-carlo-analysishttp://www.epa.gov/risk/guiding-principles-monte-carlo-analysishttp://dx.doi.org/10.1016/j.buildenv.2012.07.005http://dx.doi.org/10.1016/j.buildenv.2012.07.005http://dx.doi.org/10.1115/1.4029862http://dx.doi.org/10.2307/1912934http://dx.doi.org/10.1016/j.enpol.2015.10.004http://dx.doi.org/10.1016/j.enpol.2015.10.004http://dx.doi.org/10.1080/20430795.2016.1196556http://dx.doi.org/10.1016/j.enpol.2016.03.015http://dx.doi.org/10.1002/wene.218http://dx.doi.org/10.1002/wene.135http://dx.doi.org/10.1002/wene.135https://www.ise.fraunhofer.de/http://www.dbresearch.comhttp://www.dbresearch.com

Multivariate analysis of solar city economics: impact of energy prices, policy, finance, and cost on urban photovoltaic po...INTRODUCTIONBRIEF REVIEW OF RECENT `SOLAR CITY´ ASSESSMENT LITERATUREANALYTIC APPROACHBenefit-Cost AnalysisHard Costs Versus Soft CostsMonte Carlo SimulationRegression AnalysisModel Robustness Check

PROJECT FINANCE ANALYSISREGRESSION RESULTSRegressions by CityRegression Analyzing the Combined Pool of Cities

RESULT OF THE MODEL ROBUSTNESS CHECKCONCLUSIONNotesReferences

Multivariate analysis of solar city economics: Impact of …...risk, and reduce project costs.7,8 It...

Documents

Transcript of Multivariate analysis of solar city economics: Impact of …...risk, and reduce project costs.7,8 It...